add decimal_threshold: float = 0.01 to si_fmt()

simplify set_vmin_vmax()

Author: janosh

pymatviz/ptable.py

@@ -255,23 +255,12 @@ def data_preprocessor(data: SupportedDataType) -> pd.DataFrame:
 
     def set_vmin_vmax(df: pd.DataFrame) -> pd.DataFrame:
         """Write vmin and vmax to DataFrame metadata."""
-        # Flatten the DataFrame
-        flattened_values: list[float] = []
-
-        for item in df["Value"]:
-            for value in item:  # item is always a list
-                try:
-                    flattened_values.append(float(value))
-                except (TypeError, ValueError):  # noqa: PERF203
-                    try:
-                        flattened_values.extend(list(map(float, value)))
-                    except (TypeError, ValueError) as exc:
-                        raise TypeError(f"Unsupported data type {type(value)}") from exc
-
-        # Set vmin and vmax
-        df.attrs["vmin"] = min(flattened_values)
-        df.attrs["vmax"] = max(flattened_values)
+        # flatten up to triple nested lists
+        values = df["Value"].explode().explode().explode()
+        numeric_values = pd.to_numeric(values, errors="coerce")
 
+        df.attrs["vmin"] = numeric_values.min()  # ignores NaNs
+        df.attrs["vmax"] = numeric_values.max()
         return df
 
     # Check and handle different supported data types
@@ -295,7 +284,7 @@ def set_vmin_vmax(df: pd.DataFrame) -> pd.DataFrame:
         lambda x: np.array([x]) if isinstance(x, float) else np.array(x)
     )
 
-    # Handle missing and anomaly
+    # Handle missing and anomalous values
     data_df = handle_missing_and_anomaly(data_df)
 
     # Write vmin/vmax into metadata

pymatviz/utils.py

@@ -305,7 +305,13 @@ def pick_bw_for_contrast(
     return "black" if light_bg else "white"
 
 
-def si_fmt(val: float, fmt: str = ".1f", sep: str = "", binary: bool = False) -> str:
+def si_fmt(
+    val: float,
+    fmt: str = ".1f",
+    sep: str = "",
+    binary: bool = False,
+    decimal_threshold: float = 0.01,
+) -> str:
     """Convert large numbers into human readable format using SI prefixes.
 
     Supports binary (1024) and metric (1000) mode.
@@ -321,6 +327,11 @@ def si_fmt(val: float, fmt: str = ".1f", sep: str = "", binary: bool = False) ->
             or trailing whitespace for shorter numbers. See
             https://docs.python.org/3/library/string.html#format-specification-mini-language.
         sep (str): Separator between number and postfix. Defaults to "".
+        decimal_threshold (float): abs(value) below 1 but above this threshold will be
+            left as decimals. Only below this threshold is a greek suffix added (milli,
+            micro, etc.). Defaults to 0.01. i.e. 0.01 -> "0.01" while
+            0.0099 -> "9.9m". Setting decimal_threshold=0.1 would format 0.01 as "10m"
+            and leave 0.1 as is.
 
     Returns:
         str: Formatted number.
@@ -334,10 +345,10 @@ def si_fmt(val: float, fmt: str = ".1f", sep: str = "", binary: bool = False) ->
             if abs(val) < factor:
                 break
             val /= factor
-    elif val != 0 and abs(val) < 0.1:
+    elif val != 0 and abs(val) < decimal_threshold:
         # milli, micro, nano, pico, femto, atto, zepto, yocto
         for _scale in ("", "m", "μ", "n", "p", "f", "a", "z", "y"):
-            if abs(val) > 1:
+            if abs(val) >= 1:
                 break
             val *= factor
 

tests/test_io.py

@@ -154,12 +154,13 @@ def test_df_to_pdf(
     assert contents[:4] == b"%PDF"
 
     # Test file overwrite behavior
-    file_size_before = file_path.stat().st_size
+    file_size_before = file_path.stat().st_size  # ~7000 bytes
     df_to_pdf(**kwds)
-    file_size_after = file_path.stat().st_size
+    file_size_after = file_path.stat().st_size  # ~7000 bytes
 
     # file size should be the same since content is unchanged
-    assert file_size_before - 10 <= file_size_after <= file_size_before + 10
+    assert abs(file_size_before - file_size_after) < 2000
+    # file size difference strangely increased from <10 to 7354-6156=1198 on 2024-05-04
 
 
 def test_normalize_and_crop_pdf(

tests/test_ptable.py

@@ -84,24 +84,24 @@ def test_unsupported_type(self) -> None:
 
     def test_get_vmin_vmax(self) -> None:
         # Test without nested list/array
-        test_dict_0 = {"H": 1.0, "He": [2.0, 4.0], "Li": np.array([6.0, 8.0])}
+        test_dict_0 = {"H": 1, "He": [2, 4], "Li": np.array([6, 8])}
 
         output_df_0 = data_preprocessor(test_dict_0)
 
-        assert output_df_0.attrs["vmin"] == 1.0
-        assert output_df_0.attrs["vmax"] == 8.0
+        assert output_df_0.attrs["vmin"] == 1
+        assert output_df_0.attrs["vmax"] == 8
 
         # Test with nested list/array
         test_dict_1 = {
-            "H": 1.0,
-            "He": [[2.0, 3.0], [4.0, 5.0]],
-            "Li": [np.array([6.0, 7.0]), np.array([8.0, 9.0])],
+            "H": 1,
+            "He": [[2, 3], [4, 5]],
+            "Li": [np.array([6, 7]), np.array([8, 9])],
         }
 
         output_df_1 = data_preprocessor(test_dict_1)
 
-        assert output_df_1.attrs["vmin"] == 1.0
-        assert output_df_1.attrs["vmax"] == 9.0
+        assert output_df_1.attrs["vmin"] == 1
+        assert output_df_1.attrs["vmax"] == 9
 
 
 class TestMissingAnomalyHandle:

tests/test_utils.py

@@ -285,6 +285,8 @@ def test_si_fmt() -> None:
     assert si_fmt(-0.93) == "-0.9"
     assert si_fmt(-0.93, fmt=".2f") == "-0.93"
     assert si_fmt(-0.1) == "-0.1"
+    assert si_fmt(-0.001) == "-1.0m"
+    assert si_fmt(-0.001, decimal_threshold=0.001, fmt=".3f") == "-0.001"
     assert si_fmt(-1) == "-1.0"
     assert si_fmt(1.23456789e-10, fmt="5.1f", sep="\t") == "123.5\tp"